Sensitivity Analysis for Instrumental Variables Regression

• Activity Number: 594
• Type: Invited
• Date/Time: Thursday, August 2, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Health Policy Statistics Section
• Author(s): Jaeun Choi*+ and Mary Beth Landrum and James O'Malley
• Companies: Harvard Medical School and Harvard Medical School and Harvard Medical School
• Keywords: Comparative effectiveness research ; Instrumental variable ; Sensitivity analysis ; Observational data

Abstract

We describe a novel method for assessing the sensitivity of instrumental variables analyses to violations of assumptions in observational studies. The method is applied to the comparison of reformulated and original formulations of antidepressant medications using the likelihood a patient discontinues treatment within 90 days (a binary endpoint) and the time to discontinuation of treatment in a contemporary sample of privately insured patients (a survival time). Both area-level and individual-level instruments are considered. A novel graphical representation of the sensitivity region reveals the treatment effects consistent with the sensitivity analysis assumptions about the extent to which the IV assumptions are relaxed. If time permits, the incorporation of longitudinal data and a Bayesian approach to estimation will be discussed.